Hi Richard,
So, can an rdfs:Class also be used as a skos:Concept, or vice versa?
Yes.
OWL DL folks will obviously not like this, but that's not my problem.
A problem might be this: If there is a skos:Concept, and we use it as a
class, what exactly are the instances of that class? Having this:
foo:Cougar a skos:Concept; skos:prefLabel "Cougar".
And then doing this:
bar:Bob a foo:Cougar .
Then Bob could be an animal, or Bob could be some sort of document,
because the foo:Cougar concept may have been a "container" for documents
rather than animals. So the "bar:Bob a foo:Cougar" triple really just
tells us that bar:Bob is somehow related to or about cougars, but not
that Bob is an animal of the species Cougar. This problem could be
solved by saying that "foo:Cougar rdfs:subClassOf ex:Animal". So the
problem can be worked around. But it still shows that using concepts as
classes is tricky business.
I never pretended that was easy, just that there were fewer problems ;-)
However, I would in general expect that the link between a document and its
subject should be represented by a dedicated property that is not rdf:type [1].
If people generally adopted this pattern, there would be less chance of
someone's using the cougar concept as the class of documents about cougars,
which I expect is the most likely confusion.
But of course you can never remove all danger. When you start using a concept
as a class, I guess it would be good practice to add some documentation (e.g.
an rdfs:comment on the class) to clarify your intention.
As you say in the following, one can expect less issues for the other
direction. All (RDFS/OWL) ontologies can be regarded as knowledge organization
systems, but KOSs are generally not trivially mapped into ontologies.
The other way round -- using classes as concepts -- seems safer. I find
it hard to find a practical problem with this. Both for classes and for
concepts there are custom properties for indicating equivalence --
owl:equivalentClass and the different skos:match properties -- so one
can steer clear of owl:sameAs and avoid most of the usual coreference
problems.
Yes!
Antoine
[1] http://www.w3.org/TR/skos-primer/#secindexing
On 3 Dec 2009, at 18:58, Antoine Isaac wrote:
And in fact, while I understand that it is not very intuitive to
have Bob the cougar as a skos:Concept (even though it is technically
allowed), I see less problems for dealing this way with the class of
cougars...
Mentioning that you see less problems without going into any detail
is not exactly useful. I see more problems.
Yes, I should have been clearer.
In fact the most important argument (as I understand it from the
discussion you started at [1]) against considering a given entity as a
skos:Concept is that this entity may not have been designed as part as
a knowledge organization endeavour. It is for example rather
far-stretched to say that a person like Michelle Obama was conceived
as part of a knowledge organization system.
My point is that this is less problematic for classes, as these are
items of a knowledge organization system from the start. They are also
"taxonomist business objects", aren't they? So we could easily treat
them as skos:Concepts.
We have in fact touched this aspect in some parts of the SKOS
documentation [2,3,4]. I'd be very interested to know whether you
think this is wrong view!
Best,
Antoine
[1] http://lists.w3.org/Archives/Public/public-esw-thes/2009Nov/0000.html
[2] http://www.w3.org/TR/2009/NOTE-skos-primer-20090818/#secskosowl
[3] http://www.w3.org/TR/skos-reference/#L896
[4] http://www.w3.org/TR/skos-reference/#L1170
Best,
Antoine
2. I'm not sure if it's wise to use the same URI for the Cougar
"concept" and the Cougar "class". I don't think that this "punning"
is against any spec, but it will cause endless head-scratching
among potential users of your data. It would be more
straightforward to mint a separate URI for the class, and relating
it 1:1 to the species concept using an appropriate property
(there's probably one in UMBEL; if not, mint your own -- maybe
"speciesClass"). Since you own the URI space anyway, minting new
URIs would be cheap.
This kind of punning between concepts, things and classes is an
interesting issue, and I'm afraid that it's not yet well
understood. Avoiding it puts you on the safe side.
That being said, can you talk a bit about your motivation for
wanting to re-use the same URI?
Best,
Richard
This should work with OWL2 but I don't know how well it will work
with the
LOD.
Also I created a VERY preliminary OWL document that would contain
a much
more complete representation of the species.
My thoughts are that these OWL documents would be used to help
determine
what specimens are instances of what species concept.
The goal would be to provide an OWL document for those who need a
more
complete description of what we mean by the URI, while
also providing a much lighter RDF representation that could be
used for
concept mapping etc.
However, I don't know if I am going about this in the right way.
Below are my VERY preliminary examples of what these OWL documents
might
look like.
The example has some attributes that I thought should be included
in a
species document, but it does not have everything that would like to
eventually include.
http://rdf.taxonconcept.org/owlses/v6n7p/2009-12-01.owl
Doc's at http://rdf.taxonconcept.org/owlses/v6n7p/owl_doc/index.html
The common classes etc, would eventually be moved to a separate
ontology
that would be imported into each individual species ontology.
And these ontologies will need to be fixed so that they work
together, I
don't think they do right now.
Thanks in Advance, :-)
- Pete
----------------------------------------------------------------
Pete DeVries
Department of Entomology
University of Wisconsin - Madison
445 Russell Laboratories
1630 Linden Drive
Madison, WI 53706
GeoSpecies Knowledge Base
About the GeoSpecies Knowledge Base
------------------------------------------------------------